Regarding critiques: #1 and 5 are in direct opposition to each other. If you made this policy revenue neutral, neither would apply. You can make the policy revenue-neutral by setting an overall benchmark of safety (e.g. 2023 fatality/mile) and penalizing/rewarding in proportion to deviation from that baseline.
Regarding #2, more niche models wouldn't benefit as much from the law of large numbers.
Is there a selection bias, where more dangerous drivers are drawn to certain models (e.g. racier cars)? Given the strong brand resonance with cars, I find that likely, but I don't know how big the effect size is.
Regarding #4, that is the most obvious solution to me. Insurance should fully internalize the economics of which car models lead to what economic damages. What is it about insurance regulations that inhibit insurance companies from charging exactly enough to cover the expected damages from each car model?
On #2, yes, there is almost certainly some selection bias. e.g., you can see that certain vehicles are driven (or at least crashed) more by men than others are.
On #4: this approach assigns a dollar figure that's higher ($3 million) than most deaths wind up paying out. You could have a somewhat similar effect by saying that every time there's a car crash death, the insurance company is responsible for paying money into a fund in line with my approach above. That would create similar incentives, though it would increase the cost for drivers: insurance companies would be on the hook for about $125B more each year, and that money would have to come out of premiums. Such a program would likely add something like $500 to the cost of driving a car for a year. By contrast, I believe that my approach would be neutral for drivers cost-wise.
I personally think it's pretty reasonable to increase the cost of driving to reflect social costs, but that's not an easy political sell!
It's an interesting approach. However, the example of F-Series is tricky - it might be the highest selling car in the country, but fatalities also depend on where it's driven. If most of the F-Series cars are driven in non-urban areas, I imagine the fatalities will be low.
If we really want to do this, we will have to segment the "safety baseline" based on the environment it's targeted towards and where it's used. If a car is primarily driven in dense urban areas, its safety bar should be higher than a pickup truck. However, if we force a pickup truck to be as safe as a small sedan, it might not be able to perform its main job (being a good pickup truck) well.
We can argue as a second order effect - if Ford F-series gets lower fatalities, and it gets government subsidies, it will get cheaper and people might buy it for urban environment, resulting it's fatalities going up (as it's not actually a safer car, it was just driven in sparse environments). But now we are just playing with incentives without lowering the actual fatalaties.
We already have standardized tests for this (crash test ratings) - we should invest more in these tests so they reflect the real world better. The hypothesis would be that these ratings will effectively reduce the overall fatalaties, and help people make better decisions.
You're right that my analysis doesn't control for where vehicles are driven; a more advanced version could do that, but it complicates the analysis and I'm unsure it would be necessary.
Big Carrots and Big Sticks is a fundamentally different approach to crash tests: it argues that vehicle companies shouldn't be prohibited from manufacturing vehicles that are less safe, but that they (and ultimately the buyer of the vehicle) should bear the societal cost of doing so.
Regarding critiques: #1 and 5 are in direct opposition to each other. If you made this policy revenue neutral, neither would apply. You can make the policy revenue-neutral by setting an overall benchmark of safety (e.g. 2023 fatality/mile) and penalizing/rewarding in proportion to deviation from that baseline.
Regarding #2, more niche models wouldn't benefit as much from the law of large numbers.
Is there a selection bias, where more dangerous drivers are drawn to certain models (e.g. racier cars)? Given the strong brand resonance with cars, I find that likely, but I don't know how big the effect size is.
Regarding #4, that is the most obvious solution to me. Insurance should fully internalize the economics of which car models lead to what economic damages. What is it about insurance regulations that inhibit insurance companies from charging exactly enough to cover the expected damages from each car model?
On #2, yes, there is almost certainly some selection bias. e.g., you can see that certain vehicles are driven (or at least crashed) more by men than others are.
On #4: this approach assigns a dollar figure that's higher ($3 million) than most deaths wind up paying out. You could have a somewhat similar effect by saying that every time there's a car crash death, the insurance company is responsible for paying money into a fund in line with my approach above. That would create similar incentives, though it would increase the cost for drivers: insurance companies would be on the hook for about $125B more each year, and that money would have to come out of premiums. Such a program would likely add something like $500 to the cost of driving a car for a year. By contrast, I believe that my approach would be neutral for drivers cost-wise.
I personally think it's pretty reasonable to increase the cost of driving to reflect social costs, but that's not an easy political sell!
It's an interesting approach. However, the example of F-Series is tricky - it might be the highest selling car in the country, but fatalities also depend on where it's driven. If most of the F-Series cars are driven in non-urban areas, I imagine the fatalities will be low.
If we really want to do this, we will have to segment the "safety baseline" based on the environment it's targeted towards and where it's used. If a car is primarily driven in dense urban areas, its safety bar should be higher than a pickup truck. However, if we force a pickup truck to be as safe as a small sedan, it might not be able to perform its main job (being a good pickup truck) well.
We can argue as a second order effect - if Ford F-series gets lower fatalities, and it gets government subsidies, it will get cheaper and people might buy it for urban environment, resulting it's fatalities going up (as it's not actually a safer car, it was just driven in sparse environments). But now we are just playing with incentives without lowering the actual fatalaties.
We already have standardized tests for this (crash test ratings) - we should invest more in these tests so they reflect the real world better. The hypothesis would be that these ratings will effectively reduce the overall fatalaties, and help people make better decisions.
Thanks for your comment, Dennis.
Fatalities for crashes where large vehicles are involved are pretty high, most notably for the person/people in the _other_ vehicle:
https://mikegreenfield.substack.com/p/cars-and-trucks-and-uneven-fights
You're right that my analysis doesn't control for where vehicles are driven; a more advanced version could do that, but it complicates the analysis and I'm unsure it would be necessary.
Big Carrots and Big Sticks is a fundamentally different approach to crash tests: it argues that vehicle companies shouldn't be prohibited from manufacturing vehicles that are less safe, but that they (and ultimately the buyer of the vehicle) should bear the societal cost of doing so.